In order to meet the needs of probabilistic risk assessment of man-machine system, a quantitative method of human error probability is proposed. Skill, rule and knowledge-based (SRK) framework and Hanaman decision tree method for behavior pattern determination are analyzed. It is pointed out that it is necessary to consider the uncertainty of the influencing factors of behavior pattern in the process of behavior pattern determination; The fuzzy logic method is used to deal with the uncertainty of each influencing factor of the behavior pattern, and the fuzzy reasoning rules are constructed according to the Hanaman decision tree. The membership function of human error probability is determined by using the human error probability interval provided by systematic human action reliability procedure (SHARP) method. The results show that the uncertainty of the task scenario is considered in the method, and the accurate value of the human error probability can be obtained, which meets the needs of the probabilistic risk assessment of the man-machine system.